改进人工鱼群算法及其收敛性分析

Improved Artificial Fish Swarm Algorithm and its Convergence Analyzing

为了克服人工鱼群算法容易收敛于局部最优和解精度不高的缺点,提出了一种新的小生境人工鱼群算法(NAFS)。在算法后期根据鱼群聚集程度引入小生境排挤机制,维持种群的多样性。为了说明该算法的有效性,利用压缩映射定理从理论上证明了该算法的全局收敛性。最后,通过在四个典型Benchmark函数上的实验,并与差异演化算法、粒子群算法、鱼群算法对比,证明该算法的解精度比原始人工鱼群算法有较大的提高。

To overcome the Artificial fish swarm（AFS）＇s shortcomings what are convergenting to local best answer easily and precision of solution is lower,a new hybrid niching and fitting automatically artificial fish swarm（NAFS） is presented.In the later of running of NAFS,niching is imported according as degree of fishs＇ mass to maintain the diversity of population.To prove the ability of convergence of NAFS,compression mapping theorem is used to analyse NAFS＇s ability of the global convergence.At last,some experiments on four classical Benchmark functions and comparing with DE,AFS,PSO.It is conclusion that NAFS has higher ability in searching the global best result than original AFS.